XMM-Newton high-resolution X-ray spectroscopy of the Wolf-Rayet object WR 25 in the Carina OB1 association

Transcription

1 A&A 402, (2003) DOI: / : c ESO 2003 Astronomy & Astrophysics XMM-Newton high-resolution X-ray spectroscopy of the Wolf-Rayet object WR 25 in the Carina OB1 association A. J. J. Raassen 1,2, K. A. van der Hucht 1,R.Mewe 1, I. I. Antokhin 3,4,5,G.Rauw 3,,J.-M.Vreux 3, W. Schmutz 6,andM.Güdel 7 1 SRON National Institute for Space Research, Sorbonnelaan 2, 3584 CA Utrecht, The Netherlands a.j.j.raassen@sron.nl;k.a.van.der.hucht@sron.nl;r.mewe@sron.nl 2 Astronomical Institute Anton Pannekoek, Kruislaan 403, 1098 SJ Amsterdam, The Netherlands raassen@astro.uva.nl 3 Institut d Astrophysique et de Géophysique, Université deliège, Allée du 6 Août, 17 Bât. B5c, 4000 Liège, Belgique rauw@astro.ulg.ac.be; vreux@astro.ulg.ac.be 4 Present address: Department of Physics and Astronomy, University of Glasgow, Kelvin Building, Glasgow G12 8QQ, Scotland, UK igor@astro.gla.ac.uk 5 On leave from: Sternberg Astronomical Institute, Moscow University, Universitetskij Prospect 13, Moscow , Russia 6 Physikalisch-Meteorologisches Observatorium Davos, Dorfstrasse 33, 7260 Davos Dorf, Switzerland w.schmutz@pmodwrc.ch 7 Paul Scherrer Institut, Würenlingen & Villigen, 5232 Villigen PSI, Switzerland guedel@astro.phys.ethz.ch Received 26 November 2002 / Accepted 21 January 2003 Abstract. We report the analysis of the first high-resolution X-ray spectra of the Wolf-Rayet (WR) object WR 25 (HD 93162, WN6ha+O4f) obtained with the Reflection Grating Spectrometers (RGS) and the European Photon Imaging Cameras (EPIC-MOS and PN) CCD spectrometers on board the XMM-Newton satellite. The spectrum exhibits bright emission lines of the H- and Helike ions of Ne, Mg, Si and S, as well as Fe XVIII to Fe XX and Fe XXV lines. Line fluxes have been measured. The RGS and EPIC spectra have been simultaneously fitted to obtain self-consistent temperatures, emission measures, and elemental abundances. Strong absorption by the dense WR stellar wind and the interstellar medium (ISM) is observed equivalent to N H = cm 2. Multi-temperature (DEM) fitting yields two dominant components around temperatures of 7.0 and 32 MK, respectively. The XMM intrinsic (i.e. unabsorbed, corrected for the stellar wind absorption and the absorption of ISM) X-ray luminosity of WR 25 is L x ( kev) = erg s 1,andL x ( kev) = erg s 1, (when correcting for the ISM only) assuming d = 3.24 kpc. The obtained chemical abundances are subsolar, except for S. This may be real, but could equally well be due to a weak coupling to the continuum, which is strongly influenced by the absorption column density and the subtracted background. The expected high N-abundance, as observed in the optical wavelength region, could not be confirmed due to the strong wind absorption, blocking out its spectral signature. The presence of the Fe XXV emission-line complex at 6.7 kev is argued as being indicative for colliding winds inside a WR+O binary system. Key words. stars: individual: WR 25 stars: early-type stars: Wolf-Rayet stars: binaries: general stars: abundances X-rays: stars 1. Introduction Wolf-Rayet (WR) stars represent the one-but last phase in the evolution of massive stars with M i > 20 M.Forareviewon WR stars see, e.g., van der Hucht (1992). Send offprint requests to:a.j.j.raassen, a.j.j.raassen@sron.nl Based on observations obtained with XMM-Newton, anesascience mission with instruments and contributions directly funded by ESA Member States and the USA (NASA). Research Associate FNRS (Belgique). The first report of X-ray emission by a WR star is from Seward et al. (1979), who presented Einstein X-ray ( kev) observations of the Carina open cluster Tr 16 and its environment, including six O-type stars and one WR star, WR 25 (HD 93162, Tr16-177, WN6h+O4f; WR catalog number and spectral type from van der Hucht et al and van der Hucht 2001). Subsequent Einstein observations by Seward & Chlebowski (1982) of the same region show X-rays from 15 O-type and WR stars. The data are consistent with the hypothesis that L x L bol for all O-type stars in this region with remarkably little scatter. Yet, WR 25 had L x L bol, a factor of 30 larger than for the other WR Article published by EDP Sciences and available at or

2 654 A. J. J. Raassen et al.: XMM spectroscopy of WR 25 stars in this region: WR 22 (HD 92740, WN7ha+O9III-V) and WR 24 (HD 93131, WN6ha). Adopting all to be at a heliocentric distance of d = 2.6 kpc, they found for WR 22, WR 24 and WR 25 that L x 3.6, <2.5 and erg s 1, respectively. Einstein observations of other WR stars (Sanders et al. 1985; White & Long 1986) showed L x / L bol 10 7 with considerable scatter, but leaving WR 25 exceptionally bright in X-rays compared with WR stars of various different subtypes. Subsequent X-ray observations of WR stars by the EXOSAT, Ginga, ROSAT and ASCA satellite observatories have added considerably to the X-ray view of WR stars. For X-ray surveys of large samples of WR stars, see Pollock (1987) and Pollock et al. (1995). For reviews on X-ray properties of WR stars see, e.g., Willis & Crowther (1996), van der Hucht (2002b), and Corcoran (2003). A uniform analysis of all 48 WR stars positively detected with Einstein (Pollock1987) showedthattheirx-rayluminosities cover a range of more than two orders of magnitude. In particular: (i) single WN stars exhibit L x -values of about a factor of four larger than do single WC stars; (ii) WR+OB binary systems tend to be X-ray brighter than single WR stars; and (iii) the few WR stars with absorption lines in their spectra appear significantly more X-ray luminous than single WR stars, an indication that they may be WR+OB binaries. ROSAT observations of some 150 galactic WR stars confirm and detail this view (Pollock et al. 1995). The X-ray-brightest Galactic WR object detected to date is WR 43c (HD C, WN6ha+?, period hundreds of days, Moffat & Niemela 1984) in the cluster NGC 3603, with a Chandra-ACIS-I ( kev) unabsorbed X-ray luminosity of L x erg s 1 (Moffat et al. 2002). X-rays may be either of thermal or non-thermal origin. Assuming a thermal generation, the observed X-rays indicate temperatures of a few million degrees Kelvin. Such temperatures are not expected in the atmospheres of these hot ( kk) stars as long as radiative equilibrium holds. Thus some material must be heated by non-radiative energy transfer, e.g., by hydrodynamic shocks. In WR binaries with a massive companion these shocks may arise from colliding winds, whereas in systems with a compact companion shocks could be caused by accretion phenomena. For single WR stars, however, those shocks must be an intrinsic property of the stellar wind. According to the phenomenological model proposed by Lucy & White (1980) and further elaborated by Lucy (1982), shocks are generated throughout a radiation driven stellar wind as the consequence of dynamical instabilities. Such instabilities have been studied in detail by, e.g., Owocki & Gayley (1995, 1999), Owocki & Cohen (1999) and Dessart & Owocki (2002). Model computations predict shock velocityjumps ranging from 500 to 1000 km s 1, implying post-shock temperatures which could account for the observed thermal X-ray production. The Owocki et al. models were developed for the radiation-driven winds of OB stars, while the driving mechanism of the stronger mass-loss from WR stars is not yet established. Baum et al. (1992) have modeled the observed X-ray emission of WR stars in a semi-empirical approach, assuming a standard non-lte WR model-atmosphere component in radiative equilibrium and a hot component of shocked Table 1. Stellar parameters of WR 25. quantity value ref. spectral type WN6h+O4f 1 d (kpc) v (mag) b v (mag) M v (mag) A v (mag) E b v (mag) E B V (mag) v (km s 1 ) T (kk) 31 2 log L/L Ṁ (10 5 M yr 1 ) References: 1) van der Hucht (2001); 2) Crowther et al. (1995a). material, homogeneously distributed throughout the WR atmosphere, accounting for the free-free absorption of X-rays and their non-lte transfer. The model of Baum et al. can reproduce the low-level WR X-ray fluxes, assuming a temperature of about K and a filling factor of a few percent in terms of the mass. The observed X-rays are emerging from far out in the stellar wind, due to the large optical depths. Further modeling of X-rays from single stars is provided by, e.g., Feldmeier et al. (1997a,b), Ignace et al. (2000) and Ignace & Gayley (2002). In the case of WR binaries, Cherepashchuk (1976) and Prilutskii & Usov (1976) developed the idea that the collision of two supersonic winds in a WR+O binary system should cause a bright, extensive X-ray temperature shock to form between them. Therefore, OB+OB and WR+OB binaries do not only add X-rays generated in the individual binary components, but provide also an additional X-ray excess due to the collision of the stellar winds of the binary components (e.g., Luo et al. 1990; Stevens et al. 1992; Pittard & Stevens 1997, 2002). X-ray transitions involve the innermost atomic electrons and thus, in principle, provide a means of assessing chemical abundances, via both thermal emission-line and photoelectricabsorption-edge spectra between 0.1 and 10 kev, that is not compromised by the difficulties at longer wavelengths concerning ionization balance (Pollock 1995). They are thus of special relevance to the study of the WR stars that are generally accepted to be chemically evolved. As mentioned above, WR 25 had the most prominent WR Einstein X-ray emission excess. Also its ROSAT X-ray flux is among the larger ones for WR stars. Its X-ray luminosity excess is suggestive of a colliding-wind binary with a very long period (P > 10 yr), like WR 140 (Pollock 1989; Williams et al. 1990; van der Hucht et al. 1992; Corcoran et al. 1995). Seward & Chlebowski (1982) derived from the Einstein ( kev) data of WR 25, assuming a thermal model and d = 2.6 kpc, that L x erg s 1. Pollock (1987) reanalyzed the Einstein ( kev) data of WR 25, also assuming a thermal spectrum, of 1 kev, and d = 2.6 kpc, found that L x erg s 1. This over-estimation is due to extrapolation of the 1 kev thermal spectrum from the hard IPC band to the soft one, neglecting wind absorption (see Sect. 4.3).

3 A. J. J. Raassen et al.: XMM spectroscopy of WR Table 2. Log of WR 25 observations by XMM-Newton. revolution # 115 # 116 # 283 # 284 # 285 obs. date start [UT] 04:58 instrument RGS 1 RGS 2 MOS 1 MOS 2 PN 23:48 06:51 07: :38 integration time (hr) ROSAT ( kev) data for WR 25 yielded, again assuming d = 2.6 kpc, Lx ' erg s 1 (Pollock et al. 1995; corrected by Wessolowski 1996). ASCA (0.5 4 kev) data discussed by Skinner et al. (1995) showed no significant variability within 11 hr, and a relatively soft spectrum with a Bremsstrahlung-model fit of kt ' 1.6 kev and very little emission above 2 kev. Its derived X-ray luminosity, assuming d = 2.6 kpc, is Lx ' erg s 1. The different results demonstrate the need for a homogeneous analysis of all archive data of WR 25, as performed in Sect. 4. Basic stellar parameters of WR 25 are listed in Table 1. The binary nature of WR 25 is still a matter of debate. WR 25 combines a diluted WN6-7 emission-line spectrum (e.g., Walborn et al. 1985) with a strong early-type absorption spectrum. H. Smith (1955) assigned a WN7+O7 spectral type, confirmed by L. F. Smith (1968). Subsequently, WR 25 has been classified WN6-A by Walborn (1974), WN7+a by van der Hucht et al. (1981) and WN6ha by L. F. Smith et al. (1996), due to lack of a radial velocity solution (but this could equally well indicate either a single star status, or a pole-on binary orbit, or a very long period). For the same reason, Moffat (1978) and Conti et al. (1979) rejected a binary status, although the former noted that the absorption component in the optical spectrum of WR 25 corresponds to an O4f spectral type. Van der Hucht (2001), on the basis of the diluted emission lines in the UV spectrum of WR 25 published by Walborn et al. (1985) and the absorption-component spectral type given by Moffat (1978), provocatively settled on WN6h+O4f, the spectral type which we adopt also here. Prinja et al. (1990) determined for single WN6 stars C -wind terminal velocities averaging 1700 km s 1, while for WR 25 they find 2500 km s 1, a C wind terminal velocity common for O4 stars. Drissen et al. (1992) found optical polarization variability in WR 25 and suggested that this could be binary-induced in case of a longperiod (years) orbit. Van der Hucht et al. (1992) emphasized the correlation between excess X-ray luminosities and nonthermal radio emission for a number of long-period WR binaries. At radio wavelengths, WR 25 has been detected to date only at 3 cm (Leitherer et al. 1995; Chapman et al. 1999). Anyhow, the excess X-ray luminosity of WR 25 makes it a colliding-wind-binary candidate of considerable interest (see also Pollock 1987, 1991), and worthy of multi-frequency longterm monitoring. 1 WR Fig. 1. Top: XMM false-color image ( ) of the WR 25 field from the combined 1, 2 and exposures in revs. #283, #284, and #285. Axes correspond to RA and Dec. Energy bands selected to create this image are: kev, kev, and kev. Labeled bright sources in the field are: 1: η Car (pec); 2: HD (O3.5 V((f+ )), Walborn et al. 2002); 3: HD (O3V+O8V); and 4: HD 93129A (O2 If, Walborn et al. 2002) in the open cluster Tr 14. Numerous fainter X-ray sources are present, many of them also earlytype massive stars. Also diffuse emission is visible. Bottom: XMM 1 (rev. #284) contours, plotted over an optical Digitized Sky Survey image. The here presented XMM-Newton and observations of WR 25 allow an improvement of the determination of its Xray luminosity and a first independent abundance determination of the elements Ne, Mg, Si, S and Fe. The observed element

4 656 A. J. J. Raassen et al.: XMM spectroscopy of WR 25 Fig. 2. Top: first-order background-subtracted spectra of WR 25 of rev. 284 and 285, observed by XMM-RGS (5 38 Å) and XMM-EPIC-MOS (1 15 Å). The spectra are not corrected for the effective areas of the instruments, in order to show the different efficiencies. Several prominent lines are labelled with the emitting ions. Note the small error bars. The red line shows the best-fit model and the blue line the subtracted background. Due to the higher resolution the lines above 10 Å are more prominent in the RGS spectrum. The oxygen edge at 23 Å is invisible due to lack of flux for λ > 20 Å. Bottom: the XMM-EPIC-MOS spectrum of rev The question mark indicates an unknown emission, which happens to be located at the position of Ti K-shell lines. The cosmic Ti-abundance, however, is low and no other confirmations for the presence of Ti have been found in the spectrum. ionization stages (cf. Table 3) place constraints on the structure of the X-ray forming regions of the star, and provide tests for understanding the nature of the source of X-ray emission from this Wolf-Rayet star/binary. 2. Observations and data reduction The XMM spectra of WR 25 were recorded with the Reflection Grating Spectrometers (RGSs) and the European Photon Imaging Camera (EPIC) CCD detectors. The log of XMM observations of WR 25 is given in Table 2. For general information on XMM-Newton and its X-ray instruments, see Jansen et al. (2001), den Herder et al. (2001), Strüder et al. (2001) and Turner et al. (2001). The RGS covers the wavelength range 5 to 35 Å with a resolution of about 0.07 Å (corresponding to velocities of 4200 to 600 km s 1 ), and a maximum effective area of about 140 cm 2 around 15 Å. The first spectral order has been selected by means of the energy resolution of the CCD detectors (see den Herder et al. 2001). The data were processed with the XMM-Newton Science Analysis Software (SAS, version [5.3.3]) system.forxmm-rgs the spectrum was extracted including 95% of the cross-dispersion. The background spectrum was obtained by taking events from a region spatially offset from the source, excluding 98%. For the XMM-MOS1 the spectrum was obtained by means of extracting the events within a circle around the source with outer radius of 40.The background was subtracted by means of an annulus centered on the source with inner radius of 50 and outer radius of 64.

5 A. J. J. Raassen et al.: XMM spectroscopy of WR Table 3. WR 25 emission-line wavelengths and fluxes as observed with XMM RGS1 + 2andEPIC-MOS. Values within parentheses are the 1 σ uncertainties. RGS1 + RGS2 EPIC-MOS identification a λ(å) flux b λ(å) flux b λ(å) E (kev) ion type c Fe XXV Kα S XVI Lyβ Ar XVII He (.03) S XV He Si XIII He (.03) Si XIV Lyα 6.726(.037) 0.53(.25) (.05) Si XIII He (.028) 0.13(.07) (.04) Fe XXIV Li (.032) 0.17(.08) Mg XII Lyα 9.224(.037) 0.19(.11) (.04) Mg XI He (.050) 0.16(.11) Fe-L (.064) 0.32(.22) Ne X Lyα (.041) 0.28(.09) Ne IX He Fe XIX O (.028) 0.20(.09) Fe XVIII F1-56, 55, (.021) 0.30(.09) Fe XVII Ne (.032) 0.14(.08) Fe XVII Ne (.032) 0.14(.07) Fe XVIII F (.023) 0.19(.08) Fe XVII Ne (.069) 0.08(.07) Fe XVII Ne Fe XVII Ne1-2 Notes: a Identifications from Kelly (1987). b Observed average fluxes at Earth in 10 4 photon cm 2 s 1. No significant differences between line fluxes of rev. 284 and rev. 285 have been noticed. c For notation see Mewe et al. (1985), Phillips et al. (1999), and Note 1 of Sect A check on solar flare protons resulted in the deletions of part of the exposure time (see Table 5). Figure 1a shows a combined XMM-EPIC false-color image of the WR 25 field; Fig. 1b shows a X-ray contour diagram, derived from EPIC-MOS1 data of rev. 284, overplotted on an optical DDS image. Besides WR 25, the EPIC images reveal a number of discrete X-ray sources, most of which are associated with massive stars in the Carina complex. The Carina region harbors several very young open clusters (Trumpler 14, 15 and 16, Collinder 228 and 232) that are extremely rich in very hot and massive stars, and have varying and anomalous extinction (e.g., Thé etal. 1980; Massey & Johnson 1993; and in: Niemela et al. 1995), although some other investigators find the extinction normal (e.g., Turner & Moffat 1980; Drissen et al. 1992). Many of the objects seen in Fig. 1 were already among the first early-type stars discovered to be X-ray sources with Einstein. The observed diffusex-ray emission fromthe Carina Nebulais probably due to the combined action of the stellar winds of the earlytype stars on the ambient interstellar medium. The properties of the discrete sources and the diffuse emission in the Carina Nebula will be discussed in a forthcoming paper. 3. Spectral analysis 3.1. Emission-line identification, line fluxes Figure 2 shows a superposition of the XMM-RGS and -EPIC-MOS spectra, together with the best-fit model spectrum. In Table 3 we list the wavelengths and fluxes of the emission lines measured with the RGS and the EPIC-MOS instruments. Prominent emission lines are Fe XXV (1.87Å), S XV (5.04 Å), Si XIII (5.64 Å), Si XIV (6.20 Å), Si XIII (6.65 Å), Mg XII (8.42 Å), Mg XI (9.17 Å) in EPIC-MOS and RGS, andnex (12.13 Å), Ne IX (13.45 Å), Fe XVII (15.01 Å), Fe XVII (16.78 Å), Fe XVII (17.10 Å) in RGS-spectra only. Above this wavelength the emitted spectrum is strongly absorbed by the dense stellar wind of the Wolf-Rayet star, equivalent to N H > cm 2 (Cruddace et al. 1974). We use the term equivalent N H here, because the WR 25 WN6 wind consists for 20% of helium (Crowther et al. 1995a). In WC stars the wind consists mostly of helium, and WC+O binaries show spectacular periodic changes in the equivalent N H during their orbit, e.g., WR 140 (Williams et al. 1990) and WR 11 (Dumm et al. 2003). We note also that the winds of early-type star are ionized and the cross sections for photoelectric absorption are modified compared to neutral material (see, e.g., Waldron et al. 1998). The stronger spectral lines have been measured individually by folding monochromatic delta functions through the instrumental response functions in order to derive the integrated line fluxes. A constant background level was adjusted in order to account for the real continuum and for the pseudocontinuum created by the overlap of several weak, neglected lines. We notice that below 14 Å the spectrum is dominated by

6 658 A. J. J. Raassen et al.: XMM spectroscopy of WR 25 H-like and He-like transitions of Ne, Mg, Si, S, and above 14 Å by Fe XVII and Fe XVIII lines Global fitting and emission measure modeling Multi-temperature fitting We have determined the thermal structure and the elemental composition of WR 25 s X-ray emitting plasma by means of multi-temperature fitting and DEM-modeling to the spectrum as a whole. We fitted multi-t optically thin plasma models of the spectra (RGS+MOS)usingSPEX (Kaastra et al. 1996a) in combination with the MEKAL (Mewe-Kaastra-Liedahl) code as developed by Mewe et al. (1985, 1995). The MEKAL data base is given as an extended list of fluxes of more than 5400 spectral lines, and is available on the WWW 1. From both methods a twotemperature range of plasma activity is obtained. In the multitemperature calculations we used two temperatures which were spontaneously found by the fitting procedure. The two temperature components were coupled to two different N H absorption column densities, which were free to vary. The temperatures and the corresponding EM values are given in Table 4, together with X-ray luminosities, abundances, and statistical 1 σ uncertainties. The luminosities are model luminosities at place of emitting plasma, i.e., corrected for absorption by the ISM and by the dense stellar wind of the Wolf-Rayet star. The abundances are relative to solar photospheric values from optical studies (Anders & Grevesse 1989) except for Fe, for which we use log A Fe = (see Grevesse & Sauval 1998 and 1999) instead of 7.67 (Anders & Grevesse 1989). From Table 4 we notice that the emission from the cool component (7 MK) faces the high absorbing column density, while the emission from the hot region (33 MK) is coupled to the low N H value. This indicates that the hot component is formed higher up in the wind. The same was noticed by Pollock (2002) based on Doppler shifts and Doppler broadening of lines in the spectrum of the Wolf-Rayet binary WR140 (WC7+O4-5). The emission measure of the low temperature component is higher than that of the high temperature. The latter is highly responsible for the Fe XXV line and the hot continuum, observed by XMM-EPIC-MOS. The obtained values of observations during revs. 284 and 285 are very well comparable. No change in physical conditions have been established between the two observations Abundance determination Based on the optical spectrum Crowther et al. (1995a) derived abundance values for H/He = 4.5 (number ratio), while Crowther et al. (1995b) added that N/He = and C/He = The H/He ratio shows a H-depletion by a factor of about two compared to standard solar photospheric abundances from Anders & Grevesse (1989). This H-depletion 1 version1.10/line/ 2 Here log A Fe is the logarithm of the Fe-abundance relative to log A H = Table 4. Multi-temperature fitting for XMM spectra of WR 25. parameter rev. #284 rev. #285 d (kpc) assumed N H 1 (1021 cm 2 ) 7.7 ± ± 0.6 N H 2 (1021 cm 2 ) 2.4 ± ± 0.6 T 1 (MK) 7.1 ± ± 0.3 T 2 (MK) 33 ± 4 30± 2 EM 1 (10 56 cm 3 ) 7.4 ± ± 0.9 EM 2 (10 56 cm 3 ) 3.0 ± ± 0.2 L x [ kev] (10 32 erg s 1 ) 128 ± ± 12 Abundances a : He 2.27 b 2.27 b C 0.15 b 0.15 b N 5.9 b 5.9 b O <0.4 <0.24 Ne Mg Si S Fe Fe χ 2 red 643/ /613 Notes: L x values unabsorbed i.e., corrected for absorption by the ISM and stellar wind. a Relative to solar photospheric number abundances (Anders & Grevesse 1989 or Grevesse & Sauval 1998 and 1999 for Fe). b Fixed on literature values given in Sect was confirmed by Hamann & Koesterke (1998). No optical O- abundance is available in the literature. Due to the high N H values, resulting in strong absorption above 15 Å, no C, N, and O-lines have been measured from our X-ray spectra, and therefore no abundance values can be obtained here for those elements. For He, C, and N, the abundance values obtained by Crowther et al. (1995a,b) in the optical wavelength range have been adopted. For O only an upper limit could be determined. This value might be biased by the strong absorption. For the other elements (Ne, Mg, Si and Fe, except S) the obtained values are all subsolar. This might be real but could equally well be due to a weak coupling to the continuum, which is strongly influenced by the absorption column density and subtracted background. The relative (to each other) abundances for these elements, however, are close to solarlike values. Except for Fe, the abundance of the elements are coupled for the two temperature components. The Fefeatures, however, are strongly separated in temperature regime (Fe XVII at 7 MK only, and Fe XXV at 32 MK only). Therefore the abundances for these ions were de-coupled and different Fe-abundances were obtained for the two temperature components. These differences, however, are not significant when

7 the uncertainties in the values are taken into account. To avoid the influence of wavelength shift between the observed data and the model we checked our obtained values by determining abundances based on individual lines. No significant deviations from the values derived in the global fit occur. A. J. J. Raassen et al.: XMM spectroscopy of WR Differential emission measure (DEM) modeling To show the connectivity of the different temperature components we applied a differential emission measure (DEM) model of WR 25 s X-ray emitting plasma using the various inversion techniques offered by SPEX (see Kaastra et al. 1996b). We define the DEM by n e n H dv/dlogt or integrated over one temperature bin: EM = n e n H V,wheren e and n H are the electron and hydrogen density, respectively. In Fig. 3 we show the resulting DEM as a result from simultaneous fitting of the RGS and EPIC spectra of revolutions #284 and #285 with the regularization algorithm (top panel) and with a polynomial fit of order 8 (bottom panel) (see Kaastra et al. 1996b). We assume the same abundances as were obtained in the 2-T fit (Table 4). Although the shapes of the two methods are slightly different, the results are indistinguishable from each other in view of the statistical uncertainties. As can be seen, the emission is concentrated in two temperature intervals around 8 MK and 35 MK with total integrated emission measures of 7.1 and cm 3, respectively. The emission measures compare well with the values obtained from the multi-temperature fit. 4. A search for variability As pointed out above, WR 25 may be a long-period binary. In this case, we might expect variations of its X-ray flux and/or spectral shape with the orbital phase. In a binary system consisting of two stars with strong winds, at least part of its X flux should be produced by the wind-wind collision (Prilutskii & Usov 1976). It may display phase-locked variability either as a consequence of the changing wind opacity along the line of sight towards the shock or as a result of the changing orbital separation in an eccentric binary. In order to study this potential variability, we retrieved all available archival spectral data for WR 25 from ROSAT, ASCA (only SIS0 andsis1 data, see below), and XMM-Newton public archives. Table 5 lists the log of these observations. For completeness, we added the relevant information for our current XMM-Newton observations Data reduction As one of the primary goals of this archival study was to obtain a light curve, we had to make sure that the data were extracted and analyzed in as uniform and consistent way as possible. This includes using appropriate extraction apertures (large enough to include most of the PSF yet not to degrade signal-to-noise ratio and to avoid contamination from nearby sources; the latter especially important for the ASCA data) as well as consistent models to fit the spectra. Since the spectral characteristics and sensitivity of the three instruments are very different, the only Fig. 3. Emission measure EM (= n e n H V per logarithmic temperature bin) of WR 25 derived from the RGS and EPIC-MOS spectra in units of cm 3. Top: the regularization method. Bottom: a polynomial fit of order 8. way to get consistent fluxes is through fitting the spectra and calculating the model fluxes. We retrieved the ROSAT-PSPC screened event files from the ARNIE database at Leicester University. The source spectra were extracted from an aperture 1 in radius. The ROSAT-PSPC has an on-axis resolution of 20 (PSF-FWHM) and WR 25 is a rather isolated source that lies well inside the inner ring of the wire mesh. The background spectra were extracted from an annular region centered on the source and with an inner radius of 1 and outer radius of 2. The ASCA screened event files were retrieved from the same archive (only BRIGHT-mode data were used, as BRIGHT2- mode data represented a negligible fraction of all data available). The major problem with these data is low ASCA spatial resolution, which leads to contamination of WR 25 spectra from nearby η Car and the weaker but even more nearby O3V+O8V binary HD (see Fig. 1; Tsuboi et al. 1997, Fig. 1c). For GIS detectors, wherever one chooses the background extraction area, the background seems to be strongly contaminated by the nearby sources. This results in a GIS flux for WR 25 varying by more than 50% depending on the

8 660 A. J. J. Raassen et al.: XMM spectroscopy of WR 25 Table 5. Log of archival and present observations of WR 25. obs. instrument observation MJD a exposure total duration no. ID date (s) (s) 1 ROSAT-PSPCB rp200108n , Dec ROSAT-PSPCB rp900176n , Jun ROSAT-PSPCB rp201262n , Aug ROSAT-PSPCB rp900176a , Dec ASCA-SIS0+SIS , Aug ASCA-SIS0+SIS , Jan XMM-Newton-MOS1+PN b , Jul c XMM-Newton-MOS1+PN b , Jul c XMM-Newton-MOS1+MOS2+PN , Jun c XMM-Newton-MOS1+MOS2+PN , Jun c XMM-Newton-MOS1+MOS2+PN , Jun c XMM-Newton-RGS1+RGS2+MOS , Jun c XMM-Newton-RGS1+RGS2+MOS , Jun c a Exposure start. b MOS2 was in small window mode, with WR 25 close to the edge of the central CCD, which renders the data unusable. c Some exposure time was lost due to (solar) high soft proton rate. Average exposure time is shown here, individual exposures for MOS1, MOS2, PN, RGS1, RGS2 are slightly different. Table 6. Spectral fitting and X-ray fluxes at Earth for archival and XMM-EPIC data of WR 25. obs. N H 1 N H 2 kt 1 kt 2 χ 2 red n.p. b f x (10 12 erg s 1 cm 2 ) no. (10 22 cm 2 ) (kev) [ kev] [ kev] [ kev] / / / / / a / / / / / / / / a Signal-to-noise ratio in the hard part of the spectrum for this observation is so low that no reliable column density can be found. Thus, we fixed it at a value similar to that derived from XMM-Newton spectra. b n.p.: null hypothesis probability, i.e., the probability that purely random deviations of the data from the model would result in obtained or higher value of χ 2 red.

9 A. J. J. Raassen et al.: XMM spectroscopy of WR Table 7. Average absorbed and unabsorbed X-ray luminosities of WR 25 for the three missions ROSAT, ASCA and XMM-Newton. The absorbed X-ray luminosity is derived from the X-ray fluxes (Table 6) using the adopted distance of WR25. The unabsorbed X-ray luminosity is the X-ray luminosity of WR25 just outside the absorbing wind, thus only corrected for interstellar absorption. observatory MJD av L x (10 32 erg s 1 )ford = 3.24 kpc [ kev] [ kev] [ kev] absorbed (N H = 0) ROSAT (1-4) ± ASCA (5-6) ± ± ± 19.9 XMM-Newton (7-11) ± ± ± 1.8 XMM-Newton (12-13) ± ± ± 5.6 unabsorbed (N H = cm 2 ) ROSAT (1-4) ± ASCA (5-6) ± ± ± 10.7 XMM-Newton (7-11) ± ± ± 3.5 XMM-Newton (12-13) ± ± ± 5.9 background selection. For these reasons, we decided not to use the GIS data in the current analysis. Different CCDs of the two SIS instruments have difference responses. For this reason the source spectra for SIS0 andsis1 data were extracted from a part of a circular aperture 3 in radius lying within a single CCD frame, while the background was extracted from a rectangular area within the same CCD frame. As in the 1997 ASCA observation WR 25 is located near the edge of the field of view, the flux obtained for this observation may be unreliable. As a consistency check, we compared our results for the 1993 ASCA observation with that of Skinner et al. (1995). Our absorbed SIS1 flux in the kev band as well as model parameters are practically identical to those given by Skinner et al. The first two XMM-EPIC MOS1, MOS2andPN data sets shown in Table 5 were retrieved from the public XMM-Newton data archive. Only good events (e.g., with pattern 0 12 for the MOS, etc., see Turner et al. 2001) were considered. No indication of pile-up was found in the data. Only good-time intervals with low level of the soft proton background were included in the analysis. We adopted the most up-to-date (July 2002) redistribution matrices provided by the EPIC instrument teams and used SAS to build the appropriate ancillary response file for each observation. As the goal of this section is to get accurate estimates of the flux, we used a relatively large source extraction aperture equal to 1 in radius; the background spectra were extracted from an annulus centered on the source region (inner radius of 1, outer radius 85 ). In the first two XMM data sets from Table 5, η Car was the primary target, situated in the center of the field of view. Consequently, WR 25 was offset from the center by some 7. This may lead to systematic errors of the derived flux due to: (i) inaccuracy of the Point Spread Function of the X-ray telescopes; and (ii) a calibration error in the vignetting. According to the in-flight calibration of the EPIC-PSF (XMM report CAL-TN ), a reliable correction for the encircled energy fraction (EEF) attheoff-axis angle 7 can only be done with reasonable accuracy (better than 5% at energies below 1.5 kev for MOS1, MOS2 and 4 kev for PN). Above these limits, the error may be as large as 20% or calibration is simply non-existent (e.g., at E > 4keVforPN). For this reason we did not apply the EEF correction to the extracted fluxes in data sets This must allow one to obtain more reliable comparison between XMM fluxes of WR 25 in data sets 7, 8 and 9 11 provided that the extraction aperture is large enough. Indeed, e.g. for on-axis observations with MOS the EEF within the aperture R = 1 exceeds 92% depending on energy. We may expect a not very different fraction for the 7 offset angle. Note that while formally speaking the current parameterization of the EEF at high energies and large off-axis angles is unreliable, according to it, the difference in the EEF corrections on-axis and 7 off-axis for our aperture size does not exceed 1% for MOS1, MOS2 and 2% for PN. In data sets 12 and 13 (WR 25 on-axis) we did apply the EEF, to estimate its influence on the resulting fluxes compared to the data sets 9 11 (10, 11 are same observations as 12, 13, although different sets of instruments). As for the second source of error, we were advised by the XMM-Helpdesk that an uncertainty in the position of the optical-axis used within the SAS is currently giving an error in the flux, estimated to be about 3%, 10%, and 8% for MOS1, MOS2, and PN respectively at 7 off-axis for an energy of 4.5 kev. There is a small energy dependence on the above values with lower energies being affected less. There is also a 2 4% error introduced by the same effect on the on-axis measurements. We conclude that the error of the WR 25 absolute flux for XMM data sets 12, 13 (Table 5, WR 25 on-axis) should not exceed 5 7%. The fluxes in data sets 7 11 may be

10 662 A. J. J. Raassen et al.: XMM spectroscopy of WR 25 systematically underestimated by some 6 10% due to the lack of the EEF-correction. Apart from this systematic error, the error of the flux measured within our R = 1 aperture should not exceed 5 7% for the data sets 9 11 and may be somewhat larger for the data sets 7, Spectral shape and flux variations All data were analyzed using the XSPEC software (version ). We fixed chemical abundances at values obtained in the previous section. For XMM and ASCA data we used a twotemperature thermal plasma MEKAL model (Mewe et al. 1985; Kaastra 1992) allowing distinct column densities for both components. The column densities obtained in the best-fit models are comparable or larger than the equivalent H I column density (N H = 3.5 ± cm 2, Diplas & Savage 1994). The ROSAT sensitivity range ( kev) does not justify using a two-temperature model, so these data were fit with a singletemperature absorbed MEKAL model. The fitted model parameters are shown in Table 6 (including the reduced χ 2 value and the null hypothesis probability of the fits). Table 6 lists the model fluxes integrated over three energy ranges ( kev, kev, and kev). It can be seen indeed that the EEF-corrected fluxes in data sets 12 and 13 are about 10% higher than the fluxes in data sets On the other hand, these differences may also reflect calibration uncertainties between EPIC and RGS. The light curves in three energy bands are shown in Fig. 4 (fluxes from data sets 7 11 are plotted for XMM ). From Table 6 it is clear that the spectral shape of WR 25 has not changed much in the course of 10 years. The column density at the ROSAT epoch is somewhat larger than that at the XMM-Newton epoch, but considering the errors of the former the results are quite compatible. Also, part of the difference in N H may come from the absence of the high-temperature component in our modeling of the ROSAT data. The ASCA data (especially in the second observation) have very poor signalto-noise ratio, especially in the high energy part. This explains too good to be true χ 2 values and somewhat deviating model parameters. In the relatively short history of X-ray astronomy, the X-ray flux of WR 25 has not shown strong variability, as is evident from Table 6 and Fig. 4. The XMM fluxes obtained on time interval of about 1 year differ by some 15%. However, provided that the calibration at high energies and large off-axis angles is not very good (see above), this difference may be related to the calibration uncertainties. On the other hand, the flux differences for the soft and hard bands are quite consistent; recall that the calibration for low energies is better at large off-axis angles 3. We conclude that we have not yet found spectral shape or flux variations providing a clear indication of the suspected binarity of WR 25. This could indicate a very-long period orbit, e.g., like those of WR 146, P 550 yr, or even WR 147, P 1350 yr (Setia Gunawan et al. 2000, 2001; van der Hucht et al. 2002a), and/or a circular orbit, although even a circular orbit could 3 Note that the XMM flux errors in Table 6 and in Fig. 4 do not account for systematic vignetting errors and thus are underestimated. Fig. 4. ROSAT/ASCA/XMM X-ray light curves of WR 25. Triangles: ROSAT; squares: ASCA; and dots: XMM. Note the, relatively, very small error bars for the XMM fluxes. induce variability due to inclination-dependent line-of-sight absorption X-ray luminosity Using X-ray fluxes of WR 25 obtained in previous subsections we determined the unabsorbed luminosities of WR 25 for every mission/epoch. We did this by removing the interstellar column density NH ISM = cm 2 from the best model fits and computing the resulting model fluxes. The luminosities were then calculated from these unabsorbed fluxes for the assumed distance to WR 25 d = 3.24 kpc. To estimate the errors of the luminosities we simulated a synthetic spectrum for each instrument (ROSAT, XMM, ASCA) including the effect of photon noise, using the best fit models corrected for the ISM column density. The errors on the flux for these simulated data can then be evaluated in a standard way. The average unabsorbed X-ray luminosities of WR 25 for these three missions are shown in Table 7. Seward & Chlebowski (1982) and Pollock (1987) reported very different X-ray luminosities of WR 25 (see Sect. 1) using the same Einstein-IPC ( kev) data. The difference between these authors apparently comes from different use of the (same) data: Seward & Chlebowski used the measured count rate in the whole IPC ( kev) energy range while Pollock only used the hard band ( kev) and extrapolated the hard band count rate to the soft band assuming a 1 kev thermal model. Doing so, he neglected internal absorption in the wind of WR 25 which is evident from our XMM study. In a 1 kev thermal model without wind absorption we would expect a significant number of counts in the soft IPCband. Remarkably, though, the observed count rates are almost identical: 0.14 ± 0.02 counts/s for Seward & Chlebowski and 0.13± 0.02count/s for Pollock.

11 A. J. J. Raassen et al.: XMM spectroscopy of WR We used our best-fit model 10 (Table 6) to simulate a synthetic Einstein-IPC spectrum and then get the count rates. The simulated IPC count rate in the kev band is 0.12 counts s 1, which, in view of all uncertainties involved, is in excellent agreement with the measured value. The simulated count rate in the kev band is 0.11 counts s 1, indeed very close to the total count rate due to the above mentioned wind absorption. We conclude that the X-ray luminosity of WR 25 at the Einstein epoch is consistent with the XMM L x value. 5. The kev Fe-complex: Evidence for colliding winds and thus binarity? We measured the energy and the equivalent width of this complex by fitting a power law continuum plus a Gaussian to the spectrum extracted between 4.4 and 8.0 kev. The results are listed in Table 8. The given HJD (heliocentric Julian day) are corresponding to the middle of the exposures. No significant variability of the strength of the Fe-complex is detected between the four epochs. In Table 9 we list the observations of the 6.4 kev Feemission line complex in WR 25 and other hot massive stars. We divided Table 9 in single stars and known binaries. In fact, the temperature elevation associated with hydrodynamic shocks in the winds of single stars is usually not sufficient to produce a significant Fe Kα emission. On the other hand, in a wide binary system, the winds of the binary components collide with velocities close to v and hence the plasma in the wind interaction zone can be significantly hotter. These considerations suggest that the Fe-complex is a diagnostic for colliding-wind binaries with the Fe-complex originating in the hot plasma of the wind collision zone. Not all observations listed in Table 9 were equally sensitive at kev. And in the case of the widest listed binary, WR 147, the binary-component separation ( 417 AU, Setia Gunawan et al. 2000) may be too large to generate a visible Fe-complex in the peak of the collision cone, while the non-thermal radio emission originating in the wake of the collision cone is coming from a much larger volume. Nonetheless, Table 9 shows that the observed binaries have a larger fraction showing the Fe-complex than single stars do. Because of the small number of available data this is, of course, not statistically significant, but at least a stimulating indication. This implies that the WR object WR 110 and the O-type objects HD 93129A, HD and θ 1 Ori C could also be collidingwind binaries. Incidentally, HD 93129A has recently proven to bea60mas( 200 AU at d = 3.24 kpc) visual binary (Walborn 2003). 6. Implications of results In understanding the nature of WR 25, one very important aspect is clearly the L x /L bol value of WR 25: for the first time we are in the position to provide a meaningful and accurate luminosity and to estimate how over-luminous WR 25 really is. In fact, using the ISM-corrected luminosities and the L bol value Table 8. Properties of the Fe-complex at 6.7 kev in the XMM spectrum of WR 25. rev. HJD line energy EW # (kev) (kev) ± ± ± ± from Table 1 yields L x /L bol = (i.e., log(l x /L bol ) = 5.62). While this ratio is large, Fig. 1 in Wessolowski (1996) shows that there may be other presumably single WN stars that have similar ratios. Of course, one could argue that these other systems are less well known and may also be yet unidentified binaries. Since WR 25 is the least extreme WN star, it is also instructive to compare its L x /L bol ratio with the relation for O- type stars given by Berghöfer et al. (1997): the L x for WR 25 we derived in this study is more than a factor of 10 larger than the value expected from their relation for O-type stars. There are a number of clues about the multiplicity of WR 25: on the plus side, we have (i) the large X-ray luminosity (see above), well in excess of what is expected (even accounting for the dispersion in the empirical L x vs. L bol relations), and (ii) the high-temperature component: how could such a high temperature emission be produced in the wind of a single star? On the minus side, we have (i) the lack of short-term and long-term X-ray variability, and (ii) the lack of variabilityevidence for binarity at other wavelengths, except for the optical polarization-variability found by Drissen et al. (1992). The first counter-argument could be overcome by assuming that WR 25 is either a (very) long-period binary or a system that is nearly seen pole on. In any case, long-term monitoring of WR 25 over a broad wavelength range will be instrumental to clarify its nature. 7. Summary The object WR 25 (WN6h) has been observed by XMM-RGS and -EPIC. A broad temperature range associated with the X-ray radiation has been established by multi-temperature fitting and DEM modeling of the spectra between 5 and 60 MK with maxima around 7.5 and 35 MK. In this temperature range the total emission measure is about cm 3. Above 15 Å the radiation is absorbed by the dense wind around the WR star and no features are observed in that region. The data are consistent with two different N H values: cm 2 and cm 2 for the two temperature components. We notice that the column density for the softer component is larger than the ISM value indicating that there is significant wind absorption. No abundance anomalies have been noticed. For C and N no values could be obtained (we used fixed literature values) and all other elements were slightly subsolar (except S which was about solar). The X-ray flux and spectral shape of WR 25 as measured by ROSAT, ASCA, and XMM do not show significant